Number of Voices - Actual vs. Theoratical

[Rusty O'Hara]

Note: I guess this topic is aimed primarily at VA/digital synths et al.

Thought that I would ask your opinion on how you find the number of voices on various synths, primarily the theoretical number of voices versus the actual number of voices achieved.

I have a Virus TI for example, and according to Access:

"For starters, polyphony is more than doubled, to 80 voices under average conditions"

Which in my experience is definitely not the case, or maybe I'm not using it under average conditions. I use it live, and have had on some occasions, note stealing, or a fast bass line or arp skip a note. Not the best scenario as you can imagine.

Similar story from when I used to have a Waldorf µQ (a 25 voice version); where the literature basically states that getting 25 voices is on the basis of using a very simple patch, once you start bringing in additional functionality, 3rd Osc, effects etc, you can watch that voice count drop rapidly.

Any synths out there that you have used, where when they say "16 voices" it means "16 voices" no matter what?

And vice versa, any synths where that awesome "7,425 voices" really turned out to be a measly "2 voices"?

[phwb]

The Alesis Fusion claims to have virtually unlimited polyphony. I guess that's waht you mean by theoretical, in real world scenarios, I think the VA and FM engine has 600 voices and the sampler might have a bit more and the PM or physical modeling has 200 and something. That's what I remember, you can consult the manual or their board to be sure but basically what I'm saying is, for what these go for and what they sound like, if you have very high polyphony requirements, you should try one out? I know it's marketed as an Oasis Light which it's it's not exactly but If you just need a synth with a tone of voices it will fit the bill, and for a remarkably low price.

[sacredcow]

This is only the case for dsp-based digitals. If the oscillators are pcm then you'll never have problems like this. Probably not of much help, but it does rule a lot of earlier digitals out of this conversation.

[phwb]

I don't want to discriminate against other genres of music but to me, 8 voices is plenty for pads and anything else a monosynth will suffice but I'm an analog freak.

[shaft9000]

there's all kinds of tricks used in dsp to get things done.

In a single DSP-synth , the logic gates perform all sorts of different tasks. remember, you're not even using dedicated oscillators per se, but a % of the RAM and dsp available. Practically all software has some bugs in it.. interrupt the 'usual ' data flow it was designed to do...and things can get ugly.
Like any computer system, overloading one process can detrement the performance of another, or worse the entire synth. this is the nature of digital data signal flow, and so it goes with "virtual analog".
nothing's perfect.

[sacredcow]

Of course, I'd also like to know of a single instance in non-orchestral music where 80 notes are being used at one time.

[WDW]

Of course, I'd also like to know of a single instance in non-orchestral music where 80 notes are being used at one time.

Uh, when someone is testing the number of actual voices that a synth can play simultaneously, of course!

WD

[sacredcow]

Of course, I'd also like to know of a single instance in non-orchestral music where 80 notes are being used at one time.

Uh, when someone is testing the number of actual voices that a synth can play simultaneously, of course!

WD

I said music.

[Analog Freak]

I don't want to discriminate against other genres of music but to me, 8 voices is plenty for pads and anything else a monosynth will suffice but I'm an analog freak.

Wrong, I'm an Analog Freak! On a more serious note, I think that the longer the release on a note is the more vices you are going to need. If you use really short release on notes, it's almost impossible to run out of voices, unless you actually use more than 16 or so notes at once. Personally, I never have problems with note stealing with anything more than eight voices, but I really prefer to have at least 16 voices anyway. 32 voices is about as many as I ever see myself needing given that I loathe using a computer to sequence, I'd rather just play the notes, and I only have ten fingers. Not to offend anyone, but the whole problem of increasing voice complexity causing less voices to be available is why I don't own a VA. I prefer to have a set number of voices that I can set up however I wish. That's just my perspective though.

EDITED TO CORRECT BONEHEADED MISTAKES.

[OriginalJambo]

Erm...Analogue Freak? Don't you mean release and not sustain? If so I totally agree with you.

[memo]

The Alesis Fusion claims to have virtually unlimited polyphony. I guess that's waht you mean by theoretical, in real world scenarios, I think the VA and FM engine has 600 voices and the sampler might have a bit more and the PM or physical modeling has 200 and something.

I seriously doubt it. There's no reason they'd overspec the CPU like that. I think the maximum voices are more in the 100-180 range, depending on what synth methods are used.

For a single timbre synth, I like to have at least 8 voices of actual polyphony if I'm doing strings or pads. The JD-800 has 24 as far as its spec, but in reality it often only has 6 voices since many big patches use 4 oscillator structures. Notes will cut out unless you're careful in how you play. I prefer 12 or 16.

For multi timbre stuff, I need quite a bit. At least 24, if not 32 minimum. I tend not to use synths multitimbrally anymore - but I used to out of pure necessity.

[WDW]

Of course, I'd also like to know of a single instance in non-orchestral music where 80 notes are being used at one time.

Uh, when someone is testing the number of actual voices that a synth can play simultaneously, of course!

WD

I said music.

"Music" is all relative. One man's voice test may ba another man's 80-note chord. Good grief, those could be some tricky progressions.

WD

[meatballfulton]

This is only the case for dsp-based digitals. If the oscillators are pcm then you'll never have problems like this.

Actually most PCM based synths also have variable polyphony due to the way the voice architectures work. Motif, Fantom and Triton all utilize layering of samples....this is something that dates back to the D-50 and M1 synths of 20 years ago. Roland calls the layers partials, Korg calls them timbres, Yamaha calls them elements.

In a Motif patch there are 4 elements max each using one sample. If that sample is stereo that's 2 voices in use so the worst case for a single note being played is 8 voices (out of 128 max) being used resulting in only 16 note polyphony!!! You can reduce this even further by multitimbral layering of patches (a performance in Yamaha-speak, a combi in Korg-speak). In reality, most voices do not use four elements all at the same time , so the real polyphony is in the range of 32-64 notes most of the time. Not too many Motif patches use only one mono element, but some do...and using those will indeed let you play 128 notes at once.

The math works out pretty much the same for Korg and Roland sample-based machines...more complex patches use more voices.

[meatballfulton]

Of course, I'd also like to know of a single instance in non-orchestral music where 80 notes are being used at one time.

Glissandos on a piano 8)

[CS_TBL]

..or harp, or vibraphone, or glockenspiel, or celesta or, heck, anything with multiple strings.

Furthermore, there're good reasons to swim in oodles of voices: tiny aural details. Imagine those JV-models with 4 tones per voice, and iirc XV models have up to 8 tones per voice. Imagine you're not using layered voices, but you're manually doing all that! So, for a layer of 8 you actually have 8 MIDI tracks and 8 different instruments. While it may sound overkill, it can prove to be highly flexible when mixing. It's so easy to lower a certain tone, and to bring another one forward. You could see it as a kind of additive synthesis, but now with complete tones rather than with pure sines.

Brian Transeau springs to mind, highly layered music with oodles o' details. It's a complex and dense mix and yet it doesn't sound p00p.